Skip to content

PPO with Random Steps #316

Description

@redzhepdx

Description

Unfortunately performing random discovery steps with PPO is not working, I know PPO doesn't need this feature internally but it is still a good feature to consider for tasks with long horizons. The main issue is current_log_probs are None and the memory registering logic is not working for None values. I believe the a separate log_probability retrieval logic must be added into PPO to be able run random steps.

What skrl version are you using?

1.4.3

What ML framework/library version are you using?

Torch IsaacLab 2.0 Isaac Sim 4.5

Additional system information

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions